In Silico Study of FAK Inhibitors Containing Pyrimidine ......Mukesh Bharatbhai 1, Mahirakhatun R...

Sohanabibi RM, et al: In Silico Study of FAK Inhibitors Containing Pyrimidine...

Manipal Journal of Pharmaceutical Sciences | September 2020 | Volume 6 | Issue 2 65

AbstractA series of diaminopyrimidine derivatives were collected from various literatures and subjected to in silico 3D-QSAR (Quantitative Structure Activity Relationship) study. Particularly, a field-based 3D-QSAR by using steric, H-bond donor and acceptor, electrostatic and hydrophobic fields was performed. Field-based 3D-QSAR models were generated with a data set of 62 diaminopyrimidine derivatives using its inhibitory activity against focal adhesion kinase (FAK) receptor. The data set were comprised of the training set (45 compounds; 72%) and test set (17 compounds; 28%), which were randomly assigned. The relationship of molecular field versus inhibitory activity was established using the partial least square (PLS) method. Assessment of the model for stability and predictability was done by the leave-one-out method. The QSAR model (PLS factor 6) with good predictivity showed R² value 0.98, value 0.6321, Q² value 0.6151, RMSE value 0.6, SD value 0.1527 based on different fields.

Key words: Anticancer agents, Diaminopyrimidines, FAK inhibitors, Field-based 3D QSAR

Mirza Sohanabibi R1, Azmin M. Mogal1, Uttam A. More1, Malleshappa N Noolvi1, Patel Salman Ismail1, Vaja Mukesh Bharatbhai1, Mahirakhatun R Rana1, Payal S Jain1,3, Navdeep Singh Sethi2, Vandana Kharb4

1 Department of Pharmaceutical Chemistry, Shree Dhanvantary Pharmacy College, Kim (Surat) - 394110, Gujarat, India.

2 Department of Pharmaceutical Chemistry, Doaba College of Pharmacy, Kharar (Mohali) - 140103, Punjab, India

3 Pharmacy, Gujarat Technological University, Chandkheda, Ahmedabad-382424, Gujarat, India.,

4 Department of Pharmaceutics, Sachdeva College of Pharmacy, Ghrauan (Mohali) - 140413, Punjab, India.,

* Corresponding Author

Date of Submission: 18 Jul 2020, Date of Revision: 31 Jul 2020, Date of Acceptance: 01 Aug 2020

How to cite this article: Sohanabibi R M, M. Mogal A, More UA, N Noolvi M, Ismail P S, BharatbhaiV M, Mahirakhatun R Rana, Jain P, Sethi N S, Kharb V: In Silico Study of FAK Inhibitors Containing Pyrimidine Fragment as Anticancer Agents. MJPS 2020; 6(2): 65-78.

Research Article

In Silico Study of FAK Inhibitors Containing Pyrimidine Fragment as Anticancer AgentsMirza Sohanabibi R, Azmin M. Mogal, Uttam A. More*, Malleshappa N. Noolvi, Patel Salman Ismail, Vaja Mukesh Bharatbhai, Mahirakhatun R Rana, Payal S Jain, Navdeep Singh Sethi, Vandana Kharb

Email: [email protected]

1. Introduction“Cancer” comes from the term “karkinos” which means crab in Greek. The hallmark of cancer is the growth of dangerous, carcinogenic irregular cells within the body.1,2 During tumour growth, the cells divide and multiply progressively, forming

aggressive tumours, which may invade adjacent typical cells while continuing to spread to distant areas within the body through lymphatic circulation.3 Even though cancer may originate within a group of cells, it is often accompanied by spread to other organs resulting in tumours in the breast, lungs, liver, stomach and colorectal.4,5 Malignant growth is regarded as the main source of death in the world today.6

Malignancy is the second most normal reason for mortality after coronary illness.7 Malignancy is one of the several driving reasons for death before the age of 70 in the majority of the countries.6 According to the GLOBCAN 2018 database, there will be an estimated 18 million new cases with almost 10 million deaths in the near future.6

The different targets used to treat cancer are HER-2, Flk-1/KDR (VEGFR-2), Bcr-Abl, EGFR, Src kinase, focal adhesion kinase (FAK), farnesyl transferases (FTases), histone deacetylases (HDACs), telomerases; and signal transducers and


66 Manipal Journal of Pharmaceutical Sciences | September 2020 | Volume 6 | Issue 2

activators of transcription (STATs). Among them, FAK is one of the promising targets to develop new anticancer agents.

FAK was found by two combining lines of research in the mid ‘90s. Initially, a roughly 120 kDa protein was found as a noteworthy integrin-subordinate tyrosine phosphorylated protein. This protein was named as FAK due to its prominent localization in the focal adhesions.8,9 FAK is a significant supporter of development factor receptor and integrin signal, playing an important role in ordinary and malignant growth cells like controlling cell development and survival.10 FAK articulation and Y397 phosphorylation are over-expressed in a few advanced organized, strong malignant growths.11 It is accepted that FAK signaling is compartmentalized through communications with a variety of effector proteins that confine FAK either to central grips, endocytic vesicles, cell–cell intersections or the cell nucleus.10 The schematic representation of FAK is given in Figure 1.

Figure 1: Schematic representation of FAK

Working as a non-receptor tyrosine kinase FAK is universally communicated, directing explicit signals starting in the regions of the attachment between integrin and extracellular matrix (ECM),12 just as those activated by initiated development factor receptors.13 Assessment of human tumours has found that upgraded articulation of FAK transcripts,14 protein and expanded FAK activity 15,16 are emphatically related with metastasis, and are frequently connected with less fortunate clinical outcomes.17 Since FAK is over-expressed in several malignancies such as the tumours of the pancreas, ovaries, cervix, prostate, breast, lung, and kidney, it shows an explicable remedial potential.18,19

Until now, compounds like PF-573228, PF-562271, TAE-226, CEP-37440, and Defactinib which demonstrated reversible ATP-competitive FAK inhibition have been developed to bind to the tumour and endothelial cells.20

The present study aims to build predictive field-based 3D-QSAR model in order to gain knowledge of the structure-activity relationship for pyrimidine-based compounds as FAK inhibitors. The 3D-QSAR method was used to produce the best predictive model and this method allowed us to use various descriptors like electrostatic, hydrophobic, or steric fields, for a set of aligned ligands.

2. Materials and method3D-QSAR study for collected data set was carried out using Schrodinger suite.

2.1 DatasetUsing a data set of 62 diaminopyrimidine derivatives (Table 1) and extracting their inhibitory activity against FAK receptor (as per literature reports), field-based 3D-QSAR models were created.20-26 A training set of 45 compounds (72%) and a test set of 17 compounds (28%) were created by random assignment. The half maximal inhibitory concentrations (IC₅₀) of these 62 derivatives were converted into negative logarithmic units and represented as pIC₅₀. The relationship between the observed activity and the predicted activity was constructed using a scatter plot.

2.2. Ligands preparation and molecular alignmentThe 2D structures of the selected compounds were converted to 3D structures and energy minimization was performed by applying a force field with the help of OPLS-2003. From which, one compound having higher bioactivity with lowest energy conformation was selected as a template for molecular alignment with other compounds from dataset.27

2.3 Field-based 3D QSAR studySix 3D models were generated using various Gaussian fields, among which the one that showed maximum statistical robustness was selected and was visualized using a 3D contour map.



Ligand-targeted drug design strategies like QSAR and mapping of the pharmacophore are utilized in the process of drug discovery in several ways: activity prediction of unique molecules, searching the database for new possible molecules and important structural feature identification for activity.28 Comparative molecular field analysis (CoMFA) and comparative molecular similarity indices analysis (CoMSIA) are used to design the QSAR models. These models are the execution of the CoMFA and CoMSIA strategies with a particular arrangement of parameters. Field-based model is an arrangement subordinate strategy in which the communication vitality terms are related to biological reactions, utilizing multivariate measurable investigations. Field-based QSAR enables you to fabricate 3D-QSAR models dependent on fields, for example, electrostatic, hydrophobic, or steric fields, for a lot of adjusted ligands. Using 3D-QSAR model, the field around the ligand entity was determined.27-29

In Gaussian 3D-QSAR model collaboration, energy calculations were performed using various fields making use of Gaussian settings for field calculations. While the various fields represent the autonomic factors, the key variable is the pIC50 of the model.27 The various Gaussian contour maps are explained below.

2.3.1 Steric contour mapsThe green and yellow contour regions represent the steric field. While the green coloured sections demonstrate increased activity due to expanded mass, the yellow sections propose regions where expanded steric mass is sans any activity.

2.3.2 Electrostatic contour mapsThe blue and red contour areas demonstrate the electrostatic field, where the electronic substituents resulted in a negative to activity (represented by blue colour) while the electronegative substituents were positive to activity (represented by red colour).

2.3.3 Hydrophobic contour mapsHere, the yellow and white regions indicate the hydrophobic fields, where yellow areas are good while white areas are troublesome for activity.

2.3.4 Hydrogen bond acceptor (HBA) contour mapsThe colours red and magenta indicate the HBA fields, where red areas are good and magenta areas are troublesome for activity, respectively.

2.4.5 Hydrogen bond donor (HBD) contour mapsHere, purple, and cyan regions exhibit the HBD fields, where purple regions are positive and cyan regions are negative towards the activity, respectively.

2.4.6 Gaussian aromatic ring contour mapsIt is represented as orange and gray regions, where orange regions are good for activity and gray regions are not suitable for the activity.27, 29

2.4 QSAR model validationA leave-one-out (LOO) cross-validation strategy was used to assess the stability and predictability of the developed models. In this approach, one compound is omitted from the dataset, and using the model which is derived from the remaining compounds in the dataset, the activity of the omitted compound is predicted.27 The criteria for model selection was the robustness of the model. A test (external) set comprising of 17 compounds was used to validate the models. The root mean square error (RMSE), Q², and Pearson’s r (indicative of the predictability of the model) values were the key statistical parameters that were evaluated. Contour maps were used to demonstrate the various Gaussian fields [29].

2.4.1 Partial least square (PLS) analysisThe relationship between the molecular fields and the compound inhibitory activity was established using the PLS method. pIC₅₀ values were utilized as needy factors to produce 3D-QSAR models based on regression analysis of the autonomous factors, which comprised of the field and Gaussian forces. LOO method was used to validate the model.27

#Factors: It indicates the number of factors in the model.SD: Standard deviation of the regressionR²: Correlation coefficient of activity (experimentally observed and predicted) of the training set: Cross-validated value (LOO)R² Scramble: Average value of R² from a series of models built using scrambled activitiesStability: Effect of changes in the training set composition on the stability of model predictions



F: Ratio of varianceP: Statistical significanceQ²: Q² value of test setRMSE: Root mean square errorPearson’s r: Correlation coefficient of activity (experimentally observed and predicted) of the test setVariance in the information is evaluated by the correlation coefficient ‘r’, that estimates in what way the observed data tracks the fitted regression line. Mistakes in either the model or in the data will prompt a lack of fit. The predictive correlation coefficient is based on the test set’s molecules and determined by the following equation (1).30

7

omitted compound is predicted.27 The criteria for model selection was the robustness of the model.

A test (external) set comprising of 17 compounds was used to validate the models. The root mean

square error (RMSE), Q², and Pearson’s r (indicative of the predictability of the model) values

were the key statistical parameters that were evaluated. Contour maps were used to demonstrate

the various Gaussian fields [29].

2.4.1 Partial least square (PLS) analysis

The relationship between the molecular fields and the compound inhibitory activity was

established using the PLS method. pIC₅₀ values were utilized as needy factors to produce 3D-

QSAR models based on regression analysis of the autonomous factors, which comprised of the

field and Gaussian forces. LOO method was used to validate the model.27

#Factors: It indicates the number of factors in the model.

SD: Standard deviation of the regression

R²: Correlation coefficient of activity (experimentally observed and predicted) of the training set

𝑅𝑅𝑐𝑐𝑐𝑐2 : Cross-validated value (LOO)

R² Scramble: Average value of R² from a series of models built using scrambled activities

Stability: Effect of changes in the training set composition on the stability of model predictions

F: Ratio of variance

P: Statistical significance

Q²: Q² value of test set

RMSE: Root mean square error

Pearson’s r: Correlation coefficient of activity (experimentally observed and predicted) of the test

set

Variance in the information is evaluated by the correlation coefficient ‘r’, that estimates in what

way the observed data tracks the fitted regression line. Mistakes in either the model or in the data

will prompt a lack of fit. The predictive correlation coefficient is based on the test set’s molecules

and determined by the following equation (1).30

𝑟𝑟𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝2 = 𝑆𝑆𝑆𝑆−𝑃𝑃𝑃𝑃𝑃𝑃𝑆𝑆𝑆𝑆

𝑆𝑆𝑆𝑆 (1)

7

omitted compound is predicted.27 The criteria for model selection was the robustness of the model.

A test (external) set comprising of 17 compounds was used to validate the models. The root mean

square error (RMSE), Q², and Pearson’s r (indicative of the predictability of the model) values

were the key statistical parameters that were evaluated. Contour maps were used to demonstrate

the various Gaussian fields [29].

2.4.1 Partial least square (PLS) analysis

The relationship between the molecular fields and the compound inhibitory activity was

established using the PLS method. pIC₅₀ values were utilized as needy factors to produce 3D-

QSAR models based on regression analysis of the autonomous factors, which comprised of the

field and Gaussian forces. LOO method was used to validate the model.27

#Factors: It indicates the number of factors in the model.

SD: Standard deviation of the regression

R²: Correlation coefficient of activity (experimentally observed and predicted) of the training set

𝑅𝑅𝑐𝑐𝑐𝑐2 : Cross-validated value (LOO)

R² Scramble: Average value of R² from a series of models built using scrambled activities

Stability: Effect of changes in the training set composition on the stability of model predictions

F: Ratio of variance

P: Statistical significance

Q²: Q² value of test set

RMSE: Root mean square error

Pearson’s r: Correlation coefficient of activity (experimentally observed and predicted) of the test

set

Variance in the information is evaluated by the correlation coefficient ‘r’, that estimates in what

way the observed data tracks the fitted regression line. Mistakes in either the model or in the data

will prompt a lack of fit. The predictive correlation coefficient is based on the test set’s molecules

and determined by the following equation (1).30

𝑟𝑟𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝2 = 𝑆𝑆𝑆𝑆−𝑃𝑃𝑃𝑃𝑃𝑃𝑆𝑆𝑆𝑆

𝑆𝑆𝑆𝑆 (1)

Where SD is the sum of squared deviation between the biological activities of the test set and the training set molecules. A predictive sum of square (Press) is the sum of the squared deviation between the real and predicted activities of each compound in the test set.

A sign of the presentation of the model is acquired from the cross-validation (or prescient) r² which is characterized as equation (2).30

8

Where SD is the sum of squared deviation between the biological activities of the test set and the

training set molecules. A predictive sum of square (Press) is the sum of the squared deviation

between the real and predicted activities of each compound in the test set.

A sign of the presentation of the model is acquired from the cross-validation (or prescient) r² which

is characterized as equation (2).30

𝑅𝑅𝑅𝑅𝑐𝑐𝑐𝑐𝑐𝑐𝑐𝑐2 = 1 − ∑(Y𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝− Y𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑜𝑜𝑜𝑜𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝)²∑(Y𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑜𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑜𝑜𝑜𝑜𝑝𝑝𝑝𝑝𝑝𝑝𝑝𝑝−Y𝑚𝑚𝑚𝑚𝑝𝑝𝑝𝑝𝑚𝑚𝑚𝑚𝑚𝑚𝑚𝑚)²

(2)

Table 1: Structure of the compounds with their IC₅₀ (nM) values

Compoun

d

Chemical structure IC₅₅₀₀ (nM) Comp

ound

Chemical structure IC₅₅₀₀ (nM)

1

N

NHN NH

Cl

NH

PO

O

O

O

4.25±0.87 2

N

NHNHN

Cl

NH

O

O

PO

OO

10.69±1.4

2

3

N

NHN NH

O2N

HN

PO O

O

O

5.01±0.98 4

N

NHN NH

O2N

HN

PO N

O

O

O

5.69±0.35

5

NN

S

S

O

HNN

N

HN

HN O

Cl

N

8.57 6

NN

S

S

O

HNN

N

HN

HN O

N

316.30

(2)

Table 1: Structure of the compounds with their IC₅₀ (nM) values

Compound Chemical structure IC₅₀ (nM) Compound Chemical structure IC₅₀ (nM)

1

N

NHN NH

Cl

NH

PO

O

O

O

4.25±0.87 2

N

NHNHN

Cl

NH

O

O

PO

OO

10.69±1.42

3

N

NHN NH

O2N

HN

PO O

O

O

5.01±0.98 4

N

NHN NH

O2N

HN

PO N

O

O

O 5.69±0.35

5

NN

S

S

O

HNN

N

HN

HN O

Cl

N

8.57 6

NN

S

S

O

HNN

N

HN

HN O

N

316.30

7

NN

S

S

O

HNN

N

HN

HN O

N

F

32.70 8N

NS

S

O

HNN

N

HN

HN O

N

175.40




9

NN

S

S

O

HNN

N

HN

HN O

O

N

506.70 10

NN

S

S

O

HNN

NF

F

F

HN

HN

O N

0.07

11O NH

HNN

N

HN

ON O Cl

5.5 12 HN O

NH

N

N

FFF

NH

S OO

4

13

OHN

NH

N

N

NH O

N

N

OH

Cl

2 14

S OONN

NH

N

N

FFF

NH

HN

O

1.5

15

N

N

NH

NH

O

NH

N

N NS OO

F

FF 0.6 16

N

NCl

NH

NH

NHO

N

O

O 5.17±0.75

17

N

NCl

HN NH

HN O

O

NO

O

7.28±1.21 18

N

NCl

HN NH

HN O

O

NO

O

75.8±7.52

19

N

NCl

HN NH

HN O

O

NO

O

38±5.39 20N

NCl

HN

HN

HN O

ON

O

O

25.6±4.86

21

N

NCl

HN

HN

HN O

SN

O

O

2.58±0.83 22N

N

Cl

NH

NH

NH

HN

SO

O

O

84.6




23N

N

Cl

NH

NH

NH

HN

SO

O

O

82.5 24

N

N

Cl

NH

NH

HN

NHSO O

O

O

109.2

25

N

N

Cl

NH

NH

HN

NHSO O

F

O

O

84.7 26 N

N

Cl

NH

NH

NH

HN

SO

OF

O

86.7

27N

N

Cl

NH

NH

NH

HN

SO

OO

O

110 28N

N

Cl

NH

NH

NH

HN

SO

OO

O

113.8

29N

N

Cl

HN NH

NH

O

O

NH

SO

O

108.8 30N

N

Cl

HN NH

NH

O

O

NH

SO

O CF3

235.9

31

N

NCl

NH

NH

H2N

NHO

26±2.2 32

N

NCl

HN

HN

HN

NHO

OO

HN

NH

O 0.6±0.04

33

N

NCl

HN

HN

HN

NHO

OO

HN

NH

O 4.7±0.5 34N

NCl

HN

HNH

N

NHONN

NHN

O 2.3±0.3

35N

NCl

HN

HNH

N

NHONN

N

NHO

1.2±0.1 36

SO

O

NNH

N

N

NH

N

N

O

Cl

2.2




37

SO

O

NNH

N

N

NH O

N

N

OH

Cl

2.4 38

SO

O

NHNH

N

N

NH O

N

N

OH

Cl

9.3

39N

N

ONH

N

N

HN

S NO O

Cl

OH 3.8 40

OHN

NH

N

N

NH O

N

N

OH

Cl

3.1±1

41N

N

ONH

N

N

HNNN

Cl

OH 7.0 42N

N

ONH

N

N

HNN

N

Cl

OH 4.0

43N

N

ONH

N

N

NHO

Cl

OH 42 44N

NNH

CF3

NH N

SON 1400

45N

NHN NH

CF3

NNH

SO

O

O

OH

3.3 46 N

NHN N

CF3

NNH

SO O

O

OH

177

47N

NHN NH

CF3

NNN

O

SO

O

17 48N

NHN NH

CF3

NN

O

SO

O

3.2

49N

NHN NH

CF3

NN

O

SOO

9.2 50N

NHN N

CF3

NN

O

SO

O

8.6




51N

NHN NH

CF3

NNH

SO

O

O

OH

27 52

N

NHN NH

CF3

SHNN

O

O O

11.4

53 N

NHN NH

CF3

N NNS

O O

O

21 54N

N

CF3

HN NH

Br NNS OO

0.51

55N

NHN

CF3

NH

NBr NS OO

N

O

0.51 56N

NHN

CF3

N

NBr N

S OO

N

O

7496

57

N

N

CF3

HN NH

NNS

OO

10.2 58N

N

CF3

HN NH

NNS

OO

N

O

0.51




59N

N

CF3

HN N

NNS

OO

N

O

5.75 60N

N

CF3

HN N

NN

SO

O

N

O

14.0

61N

N

CF3

HN NH

NNS

OO

1.26 62N

N

CF3

HN NH

NNS

OO

N

O

3.21

Table 2: Actual and predicted anticancer activity of the training and test set compounds

Name

Actual

pIC₅₀Predicted

activity

Prediction

error Name

Actual

pIC₅₀Predicted

activity

Prediction

error

1 8.3716 8.3084927 -0.0631073 32 9.222 9.251785 0.0299845

2 7.971 7.9744597 0.00345968 33 8.328 8.339222 0.011322

3 8.3002 8.3012595 0.00105945 34 8.638 8.681027 0.0427268

4 8.2449 8.3513530 0.106453 35 8.921 8.892985 -0.0278147

5 8.067 7.9703879 -0.0966121 36 8.658 8.541331 -0.116269

6 6.4999 6.8265820 0.326682 37 8.619 8.701013 0.0812133

7 7.4855 8.9494700 1.46397 38 8.032 7.721774 -0.309726

8 6.756 6.8266121 0.0706121 39 8.420 8.566882 0.146682

9 6.2952 6.3001804 0.00498044 40 8.509 8.738326 0.229726

10 10.1549 10.188859 0.0339589 41 8.155 8.492636 0.337736

11 8.2069 8.2420232 0.0351232 42 8.398 8.510357 0.112457

12 8.3979 8.3494366 -0.0484634 43 7.377 7.303042 -0.0737576

13 8.699 8.7383948 0.0393948 44 5.8539 6.219256 0.365356

14 8.8239 9.15675 0.33285 45 8.4815 7.943022 -0.538478

15 9.2218 9.081907 -0.139893 46 6.752 6.669909 -0.082091

16 8.2865 8.3066936 0.0201936 47 7.7696 7.917714 0.148114



3. Results and discussionCollected all data set molecules aligned with the lowest energy conformation and highest activity compound 32 as a template (see figure 2 for the alignment of the dataset). Further in subsections, the results which are the outcome of this database alignment have been discussed.

Figure 2: Database alignment by using compound no. 32 as template

Name

Actual

pIC₅₀Predicted

activity

Prediction

error Name

Actual

pIC₅₀Predicted

activity

Prediction

error

17 8.1379 8.0491494 -0.0887506 48 8.4948 8.5562494 0.0614494

18 7.1203 7.696972 0.576672 49 8.0362 8.0538895 0.0176895

19 7.4202 7.786381 0.366181 50 8.0655 8.1586167 0.0931167

20 7.5918 7.59506638 0.00326638 51 7.5686 7.942972 0.374372

21 8.5884 8.060892 -0.527508 52 7.9431 7.9020632 -0.0410368

22 7.0726 7.0846079 0.0120079 53 7.6778 8.209767 0.531967

23 7.0835 7.327189 0.243689 54 9.2924 8.25263 -1.03977

24 6.9618 7.0307116 0.0689116 55 9.2924 9.2448936 -0.0475064

25 7.0721 7.350715 0.278615 56 5.1252 5.320476 0.195276

26 7.062 6.90403 -0.15797 57 7.9914 8.143948 0.152548

27 6.9586 7.0503882 0.0917882 58 8.3625 8.228085 -0.134415

28 6.9439 6.9972222 0.0533222 59 8.2403 7.468453 -0.771847

29 6.9634 6.804041 -0.159359 60 7.8539 7.701358 -0.152542

30 6.6273 6.5786466 -0.0486534 61 8.8996 7.83975 -1.05985

31 7.585 7.30659 -0.27841 62 8.4935 8.428397 -0.065103

*Corresponding to test set compound

Table 3: PLS statistics of field-based 3D-QSAR

# Factors SD R^2 F P RMSE Q^2 Pearson-r

1 0.7221 0.5123 65.1 2.99E-11 0.82 0.2908 0.5437

2 0.4927 0.7767 106.1 1.39E-20 0.8 0.3244 0.6006

3 0.3634 0.8805 147.4 1.23E-27 0.63 0.5853 0.7652

4 0.265 0.9375 221.2 8.68E-35 0.63 0.5795 0.7641

5 0.2017 0.9644 314.3 1.16E-40 0.6 0.6233 0.7976

6 0.1527 0.9800 464.3 1.76E-46 0.6 0.6151 0.7877

3.1 Gaussian 3D QSAR analysisTable 2 shows the data obtained from the training and test set of compounds which were used to build the model. The best model (PLS factor 6) indicated good predictive power (Table 3). The predictability of the models was assessed using the LOO method. In Gaussian field-based QSAR PLS factor no. 6 showed the best model, which had the lowest standard deviation and best R² value 0.98 with the highest F value 464.3. In case of the test set, we got the q² value as 0.6151 and Pearson-r value of more than 0.7.

3.2 Validation of the modelsThe scatter plot for the predicted activity versus the actual activity of both training as well as test set compounds is shown in Figure 3, which is based on



factor 6 of the 3D-QSAR model described above. The validated model showed strong predictive power and provided comprehensive information with regards to the structural characteristics of the compounds. Table 4 shows the Gaussian field contributions.

Figure 3: Scatter plot between predicted activity and actual activity

3.3 Contour map analysisThe correlation between the biological activity and the contour maps are shown in figures 4, 5, 6, 7, 8, and 9. The contour map for the steric fields of the compound revealed that the green colour adjacent to phenyl ring attached to pyrimidine, and nearer to the squaramide group indicated the presence of bulky groups or substituents that may be favorable for activity with the pIC₅₀ value of 9.22 in compound no 32. In compound 32, the electrostatic contours of the compound revealed that electronegative groups (red colour region) attached to squaramide group favour the activity, while blue contour near to pyrimidine ring and around the squaramide group suggest an enhanced activity due to electropositive charges. In general, a pyrimidine ring in all the compounds is highly responsible for establishing hydrogen bonding with the Cys 502. Red contour

adjacent to the N-methylbenzamide, squaramide group and around the pyrimidine ring indicates that the presence of hydrogen bond acceptor subunits increases the activity. The HBD contour map of the compound suggests that purple contour around the diphenylpyrimidine ring and N-methylbenzamide group favoured the activity, while the distant cyan contour indicates that they are highly active. The hydrophobic contour map of the compound suggests that yellow colour contour around the phenyl ring of N-methylbenzamide moiety and substituent attached to squaramide group has better activity. In compound 32, the aromatic ring contour map suggests that orange contour around the N-methylbenzamide favour the activity which is present in almost all the compounds.

Figure 4: Steric contour map of compound no. 32

Figure 5: Electrostatic contour map of compound no. 32

Table 4: Field contributions of Gaussian fields

# FactorsGaussian

steric

Gaussian

electrostatic Gaussian hydrophobic

Gaussian H-bond

acceptor

Gaussian H-bond

donor

1 0.322 0.095 0.227 0.191 0.165

2 0.285 0.112 0.233 0.207 0.164

3 0.266 0.116 0.242 0.221 0.155

4 0.266 0.115 0.249 0.213 0.157

5 0.266 0.115 0.253 0.212 0.154

6 0.269 0.115 0.256 0.213 0.147



Figure 6: Hydrogen bond acceptor contour map of the compound no. 32

Figure 8: Hydrophobic contour map of compound no. 32

4. Structure activity relationship (SAR)

On the basis of 3D-QSAR contour map analysis, we came up with the SAR figure. The structure activity analysis showed that 2, 4-diaminopyrimidine is the basic scaffold for activity against focal adhesion kinase. Substitution on R1 and R2 position with alkyl chain may decrease the activity. Aromatic and heterocyclic ring substitution at R1 and R2 position may enhance the activity. Also, at C5 position of pyrimidine fragment, electron withdrawing group like Cl and NO2 are essential for activity.

Figure 10: Structure activity relationship of compound no. 32

Figure 7: Hydrogen bond donor contour map of compound no. 32

Figure 9: Aromatic ring contour map of compound no. 32



5. ConclusionIn silico study was carried out for the database of compounds collected from the different literature sources. A Gaussian-based 3D-QSAR model was developed which demonstrated satisfactory predictability and indicated the potential structural characteristics which influenced the FAK inhibitory activity. The developed QSAR model was robust and showed R²₌ 0.98, ₌0.6321, Q²₌ 0.6151, RMSE₌ 0.6, SD₌ 0.1527 based on different fields. The contour map analysis revealed that the aromatic substituent at C2 and C4 positions of diaminopyrimidine enhances the activity, while alkyl groups are not that favorable for the activity. Similarly, an aromatic ring with electronegative substituent at C2 position of diaminopyrimidine and an aromatic ring with electropositive groups at C4 position of diaminopyrimidine may enhance the activity. Notably, all the diaminopyrimidine derivatives from the database was shown to form three strong hydrogen bonds. These hydrogen bonds were formed due to the bonding between, the nitrogen atom of the pyrimidine fragment and amino acid of Cys502, the amino acid of SP564 and the carbonyl group of N-methylbenzamide, and the nitrogen atom of N-methylbenzamide with Arg505, Gly506, and water. In addition, C-5 substituent of the pyrimidine fragment, including the NO2 and Cl groups produced equivalent interaction forces with the amino acid Gln438. At C-5 position, chlorine atom produces van der waal contacts with the amino acid Met499. Due to the similar binding forces, most of the inhibitors from the dataset displayed potent inhibitory potency against FAK kinase activity.

Acknowledgements

This study is financially supported by Research Promotion Scheme, All India Council for Technical Education, New Delhi, file no. 8-246/RIFD/RPS(POLICY-1)2018-19. The authors would like to thank Dr N D Jivani, President, Dr Dinesh L Sutaria, Secretary, Shree Dhanvantary Pharmacy College for providing necessary facilities to complete this work.

References1. Long ER. A History of Pathology. Dover

Publications; 1965.

2. Schlotterbeck ME. What Is Cancer? The American Cancer Society invites all nurses to see and use this new teaching film. Am. J. Nurse 1949; 49(5):300.

3. Rajpal S, Kumar A, Joe W. Economic burden of cancer in India: Evidence from cross-sectional nationally representative household survey, 2014. PloS one 2018; 13(2): e0193320.

4. Joseph F, Fraumeni J. Epidemiologic approaches to cancer etiology. Annu. Rev. Public Health 1982; 3:80-100.

5. World Health Organization. Cancer control: knowledge into action, WHO guide for effective programmes. WHO. 2006; Geneva. https://www.who.int/cancer/modules/en/

6. GLOBOCAN-2018, “International agency for Research on cancer”, http://www.iarc.fr/index.php

7. Jemal A, Siegel R, Ward E, Murray T, Xu J, Thun MJ. Cancer statistics. CA: Cancer J. Clin 2007; 57:43-66.

8. Guan JL, Trevithick JE, Hynes RO. Fibronectin/integrin interaction induces tyrosine phosphorylation of a 120-kDa protein. Cell Regul 1991; 2(11):951–964.

9. Kornberg LJ. Signal transduction by integrins: increased protein tyrosine phosphorylation caused by clustering of beta 1 integrins. Proc. Natl Acad. Sci. USA 1991; 88 (19):8392–8396.

10. Kleinschmidt EG, Schlaepfer DD. Focal adhesion kinase signaling in unexpected places. Curr. Opin. Cell Biol 2017; 45:24-30.

11. Sulzmaier FJ, Jean C, Schlaepfer DD. FAK in cancer: mechanistic findings and clinical applications. Nat. Rev. Cancer 2014; 14:598–610.

12. Frame MC, Patel H, Serrels B, Lietha D, Eck MJ. The FERM domain: organizing the structure and function of FAK. Nat Rev Mol Cell Biol 2010; 11:802-814.

13. Chen HC, Chan PC, Tang MJ, Cheng CH, Chang TJ. Tyrosine phosphorylation of focal adhesion kinase stimulated by hepatocyte growth factor leads to mitogen-activated protein kinase activation. J Biol Chem 1998; 273:25777-25782.

14. Weiner TM, Liu ET, Craven RJ, Cance WG. Expression of focal adhesion kinase gene and invasive cancer. Lancet 1993; 342:1024-1025.

15. Owens LV, Xu L, Craven RJ, Dent GA, Weiner TM, Kornberg L, Liu ET, & Cance WG.



Overexpression of the focal adhesion kinase (p125FAK) in invasive human tumors. Cancer Re, 1995; 55:2752-2755.

16. Hess AR, Postovit LM, Margaryan NV, Seftor EA, Schneider GB, Seftor RE, Nickoloff BJ, & Hendrix MJ. Focal adhesion kinase promotes the aggressive melanoma phenotype. Cancer Res 2005; 65:9851-9860.

17. Pylayeva Y, Gillen K M, Gerald W, Beggs HE, Reichardt LF, & Giancotti FG. Ras- and PI3K-dependent breast tumorigenesis in mice and humans requires focal adhesion kinase signaling. J Clin Invest 2009; 119:252-266

18. Zhang J, He DH, Zajac-Kaye M, Hochwald SN. A small molecule FAK kinase inhibitor, GSK2256098, inhibits growth and survival of pancreatic ductal adenocarcinoma cells. Cell cycle. Pharmacol. Ther 2014; 13(19): 3143-3149.

19. Golubovskaya VM. Focal adhesion kinase as a cancer therapy target. Anti-cancer Agents Med. Chem 2010; 10; 735-741.

20. Su Y, Li R, Ning X, Lin Z, Zhao X, Zhou J, Liu J, Jin Y, Yin Y, “Discovery of 2, 4-diarylaminopyrimidine derivatives bearing dithiocarbamate moiety as novel FAK inhibitors with antitumor and anti-angiogenesis activities.” Eur. J. Med. Chem., 2019, 177, 32-46.

21. Liu H, Wu B, Ge Y, Huang J, Song S, Wang C, Yao J, Liu K, Li Y, Li Y, Ma X. Phosphamide-containing diphenylpyrimidine analogues (PA-DPPYs) as potent focal adhesion kinase (FAK) inhibitors with enhanced activity against pancreatic cancer cell lines. Bioorg. Med. Chem. Lett 2017; 25(24):6313-21.

22. Wang L, Ai M, Yu J, Jin L, Wang C, Liu Z, Shu X, Tang Z, Liu K, Luo H, Guan W. Structure-based modification of carbonyl-diphenylpyrimidines (Car-DPPYs) as a novel focal adhesion kinase (FAK) inhibitor against various stubborn cancer cells. Eur. J. Med. Chem 2019; 172:154-62.

23. Qu M, Liu Z, Zhao D, Wang C, Zhang J, Tang Z, Liu K, Shu X, Yuan H and Ma X. Design, synthesis and biological evaluation of sulfonamide-substituted diphenylpyrimidine derivatives (Sul-DPPYs) as potent focal adhesion

kinase (FAK) inhibitors with antitumor activity. Bioorg. Med. Chem 2017; 25(15):3989-3996.

24. Yen-Pon E, Li B, Acebron-Garcia-de-Eulate M, Tomkiewicz-Raulet C, Dawson J, Lietha D, Frame MC, Coumoul X, Garbay C, Etheve-Quelquejeu M, Chen H. Structure-based design, synthesis, and characterization of the first irreversible inhibitor of focal adhesion kinase. ACS Chem. Biol 2018; 13(8):2067-2073.

25. Ott GR, Cheng M, Learn KS, Wagner J, Gingrich DE, Lisko JG, Curry M, Mesaros EF, Ghose AK, Quail MR, Wan W. Discovery of clinical candidate CEP-37440, a selective inhibitor of focal adhesion kinase (FAK) and anaplastic lymphoma kinase (ALK). J. Med. Chem 2016; 59(16):7478-7496.

26. Farand J, Mai N, Chandrasekhar J, Newby ZE, Van Veldhuizen J, Loyer-Drew J, Venkataramani C, Guerrero J, Kwok A, Li N, Zherebina Y. Selectivity switch between FAK and Pyk2: Macrocyclization of FAK inhibitors improve Pyk2 potency. Bioorg. Med. Chem. Lett 2016; 26(24):5926-5930.

27. Yadav DK, Sharma P, Misra S, Singh H, Mancera RL, Kim K, Jang C, Kim MH, Pérez-Sánchez H, Choi EH, Kumar S. Studies of the benzopyran class of selective COX-2 inhibitors using 3D-QSAR and molecular docking. Arch. Pharmacal. Res 2018; 41(12);1178-1189.

28. Kalva S, Vinod D, Saleena LM, “Field-and Gaussian-based 3D-QSAR studies on barbiturate analogs as MMP-9 inhibitors. Med. Chem. Res 2013; 22(11):5303-5313.

29. Joshi SD, More UA, Dixit SR, Korat HH, Aminabhavi TM, Badiger AM. Synthesis, characterization, biological activity, and 3D-QSAR studies on some novel class of pyrrole derivatives as antitubercular agents. Med. Chem. Res 2014; 23(3):1123-1147.

30. Gade DR, Makkapati A, Yarlagadda RB, Peters GJ, Sastry BS, Prasad VR. Elucidation of chemosensitization effect of acridones in cancer cell lines: Combined pharmacophore modeling, 3D QSAR, and molecular dynamics studies. Comput. Biol. Chem 2018; 74:63-75.

In Silico Study of FAK Inhibitors Containing Pyrimidine ......Mukesh Bharatbhai 1, Mahirakhatun R...

Documents

Transcript of In Silico Study of FAK Inhibitors Containing Pyrimidine ......Mukesh Bharatbhai 1, Mahirakhatun R...